What is FLUX 3?

Last updated: August 26, 2026

What is FLUX 3?

FLUX 3 is our video generation model: one multimodal model that creates video with synchronized audio. Speech with lipsync, sound effects, and ambience are generated together with the frames, in one request.

Key Capabilities

  • Video with synchronized audio - Multilingual speech with strong lipsync, effects, and ambience, generated with the frames

  • Up to 20 seconds per request - At hd or fhd resolution (fhd is 1920 × 1088 for 16:9), 24 fps

  • Multiple scenes in one generation - Shots hold together across cuts and camera angles

  • Stylistic range beyond cinematic - Animation, motion design, and stylized artistic looks

  • Accurate text and typography - Rendered inside the scene

Modes

Every request names a mode, which tells FLUX 3 what you are starting from. All modes run on the same flux-3-video endpoint.

Mode

What it does

You provide

Text to Video (t2v)

Turns your prompt into a clip, sound included.

A prompt

Image to Video (i2v)

Your images become frames of the video: one starts the clip, two start and end it, up to ten total.

A prompt and 1-10 keyframe images

Video Continuation (v2v)

Continues an existing clip from its final frames.

A prompt and a video

Draft mode

Set draft: true to generate a fast, cheaper preview of a clip. The result includes a download URL for a draft bundle: pass that bundle back later (mode draft_enhance) to render the full-quality version of the same generation. Explore variants freely and only pay full price for the shot you keep.

Options

  • Duration - Any whole second from 5 to 20 (5 to 15 for Video Continuation), or auto

  • Aspect ratio - auto, 21:9, 2:1, 16:9, 4:3, 1:1, 3:4, or 9:16

  • Resolution - hd, or fhd for a higher-resolution result

  • Audio - On by default; set generate_audio: false for a silent clip

Pricing

FLUX 3 is priced per second of output video: full renders from $0.17/s and drafts from $0.06/s. See the pricing article for the full per-mode table.

Full documentation, including request examples for every mode, is in the FLUX 3 section of docs.bfl.ai.